1. Field of the Invention
The present invention relates to a spectral interferometry apparatus and method, which can be used to supply unambiguous profiles of the reflectivity versus optical path difference and recognise positive from negative values of optical path difference.
2. Description of the Related Art
There is a growing interest in the application of low coherence interferometry in the general field of sensing. Low coherence interferometry methods provide absolute distance measurements and are well suited for imaging rough reflecting surfaces or producing slices in the volume of diffusive and scattering media. There are different methods which obtain depth resolved information using low-coherence optical sources, and one such method uses dispersion of the spectrum. The periodicity of the channeled spectrum is proportional to the optical path difference (OPD) in an interferometer, as described as long ago as 1837, as the so called “curious bands of Talbot”. Recent presentations of such an old phenomenon are reported by A. L. King and R. Davis in “The Curious Bands of Talbot” published in the American Journal of Physics, vol. 39, (1971), p. 1195-1198 and by M. Parker Givens, “Talbot's bands”, American Journal of Physics, 61, (7), (1993), p. 601-605.
Channeled spectrum methods have been used in the sensing and fibre optic sensing field. Recent implementations have used photodetector or CCD arrays to display the channeled spectrum, as disclosed in “Channeled Spectrum Display using a CCD Array for Student Laboratory Demonstrations”, published by A. Gh. Podoleanu, S. Taplin, D. J. Webb and D. A. Jackson in the European J. Phys., 15, (1994), p. 266-271.
Channeled spectrum has also been employed in a method called “spectral optical coherence tomography” (SOCT), as disclosed in “Coherence Radar and Spectral Radar—New Tools for Dermatological Diagnosis”, published by G. Hausler and M. W. Lindner, in J. Biomed Optics, January 1998 D, Vol. 3 No. 1, pp. 21-31 and disclosed in the following patents: U.S. Pat. No. 4,932,782, Channeled light spectrum measurement method and device, P. Graindorge; U.S. Pat. No. 5,317,389; Method and apparatus for white-light dispersed-fringe interferometric measurement of corneal topography, Hochberg et al. Further such methods have been disclosed in U.S. Pat. No. 6,072,765, Optical Disk Readout Method using Optical Coherence tomography and Spectral Interferometry, by J. P. Rolland and P. J. Delfyett. The advantage of the spectral methods is that the OPD information is translated into the periodicity of peaks and troughs in the channeled spectrum and no mechanical means are needed to scan the object in depth, in for example, optical coherence tomography (OCT) of tissue. Furthermore, no mechanical means are needed to explore the OPD in multiplexed sensor arrays in such methods. If multi-layered objects are imaged, such as tissue, each layer will imprint its own channeled spectrum periodicity, depending on its depth, with the amplitude of the spectrum modulation proportional to the square root of the reflectivity of that layer. A fast Fourier transform (FFT) of the spectrum of a charge coupled device (CCD) signal translates the periodicity of the channeled spectrum into peaks of different frequency, with the frequency directly related to the path imbalance. Such a profile is termed as an A-scan in OCT, i.e. a profile of reflectivity in depth.
A possible bulk implementation of a prior art SOCT apparatus is shown in FIG. 1. In this arrangement, an optical beam from a source 1 is collimated by a collimating element 2, which could be a simple lens or achromat, or a mirror or combination of lenses or mirrors, to form the beam 3. The beam 3 is then directed towards a beam-splitter 4. The source 1 is broadband and may be for example one or more light emitting diodes, superluminescent diodes, bulb lamps or short-pulse lasers combined to produce the largest possible bandwidth and minimum spectrum ripple by techniques known in the art. The source 1 has a central wavelength suitable for the particular object to be investigated. For the investigation of a patient's eye using OCT, a wavelength in the near infrared, such as 800 to 900 nm is used. For examining skin, a wavelength of 1300 nm may be used. For sensing applications, wavelengths in the telecommunication band of 1500 nm are preferably used.
The light received by the beam-splitter 4 is split into a first optical path 41 leading to a mirror object 5, and into a second optical path 42 leading to a reference mirror 6. After reflection on the two mirrors 5 and 6 and after crossing the beam-splitter 4, the resulting two beams are superposed on an optical spectrum dispersing means, 7, for spectral analysis. The optical spectrum dispersing means 7 could be one or more diffraction gratings, one or more prisms, or combination of prisms or gratings. In the optical spectrum dispersing means 7 the spectrum is dispersed (when using a prism or prisms) or diffracted (when using a diffraction grating or gratings), and a fan of rays with different wavelengths is output. This is subsequently focused by a focusing element 8 onto a reading element, a linear photodetector array or a CCD linear array 9. An electrical spectrum analyser 91 provides the FFT of the signal delivered by the reading element 9.
The distance from the beam splitter 4 to the mirror 5 is denoted by Z. However, in other prior art arrangements, in which the mirror is replaced by a thick scattering, multi-layer object, Z is the distance from the beam splitter 4 to a scattering point or layer within the object. This means that the object path is 2Z. The distance between the beam splitter 4 and the mirror 6 is X, which means that the length of the reference path is 2X.
The channeled spectrum periodicity depends on the OPD, defined as:OPD=2(Z−X)
Consider the arrangement in which the mirror 5 is replaced by a thick scattering multi-layer object. In this case, as the periodicity depends on the modulus of the OPD, scatterers or layers symmetrically placed around the position at which the OPD is zero give the same periodicity in the channeled spectrum. If Fourier transformed, besides the terms corresponding to the useful range of OPDs, symmetrically placed terms are obtained, often referred as mirror terms and a problem associated with Fourier domain OCT can be termed as the problem of mirror terms. This introduces errors in the depth profile of the OCT system used for imaging. Equivalently when channeled spectrum is used for sensing, there is a cross-talk of signals from sensors placed at the same value of OPD either side of the zero point. Therefore, all the prior art spectral (Fourier domain) OCT methods discussed above rely on an adjustment of the object position in such a way that the scatterers in the depth of the object are confined within a single sign range of OPDs, i.e. either positive or negative. Such an adjustment complicates the measurement procedure, and may not be applicable in all situations.
For the purposes of this description, the OPD in the interferometer will be defined as the Object Path Length minus the Reference Path Length. For example, if the object to be examined is the retina, then the origin of OPD could be set somewhere in the vitreous, in front of the retinal nerve fibre layer. This will mean that the retina scatterers are all at positions such that the OPD is greater than zero. However, if the vitreous has defects within the same path range, then these defects will appear in the final depth profile of the retina. Thus, a need exists for procedures to eliminate the peaks outside the interesting range, or to make the system sensitive to the sign of the OPD.
A method called “phase shifting spectral interferometry” has recently been introduced to eliminate the terms for one sign of the OPD range, in order to address the problem of mirror terms. By introducing exact phase shifts between the two optical interferometer paths of successive CCD frames, and combining the spectra collected, it is possible to reduce the noise as well as eliminate one of the autocorrelation terms in the electrical Fourier transform spectrum of the CCD signal (for positive or negative OPD). The method allows correct reconstruction of layers in depth. However, phase shifting spectral interferometry has the following disadvantages. The phase shifts have to be accurate to within a few degrees, which requires precise control of the movement of the reference mirror. Also, because the final spectrum is delivered only after at least a number M of spectra are collected, the process is M times slower than conventional methods. A method of phase shifting spectral interferometry using five steps was disclosed in: “Fourier-domain optical coherence tomography: next step in optical imaging”, by M. Wojtkowski, A. Kowalczyk, P. Targowski, I. Gorczynska, published in Optica Applicata, Vol. XXXII, No. 4, (2002), p. 569-580. When using this method, five frames are required before providing an OCT image. However, the most important disadvantage associated with phase shifting spectral interferometry is movement of the object being examined, for example tissue. Movement of the tissue being examined alters the value of the phase shift and has the effect of bringing back the terms for the sign of OPD (i.e. positive or negative) which otherwise would have been cancelled by the phase shifting method.
The paper entitled “Theoretical Study of Talbot-like Bands Observed Using a Laser Diode Below Threshold”, by A. Gh. Podoleanu, S. Taplin, D. J. Webb and D. A. Jackson, published in J. Pure and Applied Optics, Vol. 7, (1998), pp. 517-536 and “Talbot-like Bands for Laser Diode Below Threshold”, by A. Gh. Podoleanu, S. Taplin, D. J. Webb, D. A. Jackson, published in J. Pure and Applied Optics, vol. 6, issue 3, (1997), pp. 413-424, both report about Talbot bands using laser diodes below threshold levels. The latter paper also introduces a modified Michelson interferometer and such an apparatus will now be described with reference to FIG. 2.
FIG. 2 shows a similar arrangement to FIG. 1, but with the addition of two screens 20 placed in the optical paths 41 and 42. The two screens are arranged to block out half the diameter of the optical beams 41 and 42. Explanation of operation of the set-up in FIG. 2 will be provided for the case when the dispersing means 7 is a diffraction grating.
Consider the situation in which the beam reaching the diffraction grating 7 covers N grating lines. By introducing two screens 20, halfway through into the two optical paths, spatial separation of the two beams 41′ and 42′ occurs. The beams 41′ and 42′ are what is left out of the beams 41 and 42 after passing through the beam-splitter 4.
Usually, for maximum visibility of the interference result, those skilled in the art of interferometry understand that the height of the object beam 41 has to be adjusted to be at the same height as the reference beam 42. This can be achieved by conveniently tilting the beamsplitter 4, mirrors 5 and 6, to cause the beams 41 and 42 after reflection on mirrors 5 and 6 to be at the same height with the incoming beam 3. The beams 3, 41 and 42 are in the plane of the drawing. After introducing the two screens 20 into the two optical paths, the resulting beams 41′ and 42′ are parallel and relatively displaced in a displacing plane which in this particular case is identical with the drawing plane. The line connecting the centres of the two displaced beams 41′ and 42′ drawn in a direction perpendicular to the two beams is perpendicular to the grating lines.
The election in OPD can be explained by considering the two beams output of the interferometer as comprising a number of wavelets equal to the number of grating lines excited. In the arrangement shown in FIG. 2, the screens 20 are introduced halfway through the diameter of the beam and N/2 lines are excited instead of N lines corresponding to the whole beam diameter, therefore each wave-train comprises N/2 wavelets. As consequence of the Bragg grating equation applied for the first diffraction maximum, there is a delay of λ between each wavelet and its neighbour in the wave-train. This means that each wave-train is Nλ/2 long. Due to the action of the two screens 20, the two beams 41′ and 42′ are laterally displaced by a half-diameter of the initial beam. In the same way there is a delay of λ due to the Bragg equation from a grating line to the next grating line, there is an intrinsic initial delay of Nλ/2 between the two wave-trains diffracted, because the half diameter covers N/2 grating lines. Therefore, for the condition that the OPD is zero in the interferometer, the two wave-trains of length Nλ/2 after diffraction incur an intrinsic delay of Nλ/2. This means that their overlap is zero, which results in the channeled spectrum visibility being zero. By increasing the OPD in the interferometer, the wave-trains will overlap which results in a channeled spectrum. The minimum measurable OPD is the coherence length of the source, LC, when at least two peaks are generated in the channeled spectrum. The overlap of the two wave-trains is maximum when the OPD in the interferometer equals Nλ/2, i.e. when the wave-trains are delayed by Nλ/2. In other words, the OPD in the interferometer has totally compensated for the intrinsic delay. It will readily be apparent to those skilled in the art that the overlap of the wave-trains reduces again when the OPD in the interferometer is larger, with the overlap reduced to zero when the wave-trains are delayed by their length in top of the intrinsic delay, i.e. for a total delay of Nλ which gives the maximum OPD range.
The OPD created in the interferometer is not sufficient in order to explain behaviour of the apparatus in FIG. 2. The OPD created in the interferometer combines with the intrinsic delay between the two laterally displaced beams, Nλ/2, however, the channeled spectrum periodicity corresponds to the OPD in the interferometer only.
As explained in Podoleanu's papers mentioned above, delaying the two sides of the beam propagating to the grating results in a channeled spectrum when the OPD in the interferometer has a particular sign only. These papers make distinction between two cases, designated as L and R.
In the L case, the angles at which the diffraction grating 7 is used and the position of the screens 20 are such that the component of the reference beam 42′ after diffraction is delayed by Nλ/2 behind the wave left from the object beam 41′ after diffraction. This means that an OPD in the Michelson interferometer produces a channeled spectrum and modulation of the CCD photodetector signal as long as it is between zero and Nλ.
In the R case, the angles at which the diffraction grating is used and the position of the screens are such that the wavetrain of the beam 41′ after diffraction is delayed by an intrinsic delay Nλ/2 behind the wavetrain of the beam 42′ after diffraction. This means that an OPD in the Michelson interferometer produces modulation of the CCD photodetector spectrum as long as it is between zero and −Nλ.
If the screens 20 are introduced into the beams 41 and 42 from the other side, then the beams 41′ and 42′ change their position after the beam-splitter 4 and the behaviour of the system changes from the case R to L and vice versa.
Similar explanations can be provided for other orders of diffraction or for a prism based spectral analysing element. In fact, the Talbot bands have been observed using a prism. When using the prism, the two incident beams are parallel and in a plane defined by the normal to the incident surface and the prism bisectrix. The fans of dispersed rays from both beams are contained in the same plane, defined by the normal to the exit surface and the prism bisectrix.
This is the key element in implementing a spectral OCT which can produce correct A-scans even if the origin of OPD in the interferometer is within the tissue. This is also the key element in selecting sensors in a multiplexed array by spectral low coherence interferometry, depending on the OPD corresponding to each sensor. However, the implementation described in the Podoleanu's papers above reduces the power of the signal two times in each beam due to the presence of the screens. Secondly, the low coherence sources are very sensitive to optical feedback and the Michelson interferometer returns light back to the source. Thirdly, the method of modifying the wave-train lengths in the two beams using the screens is inefficient, and the power is dependent on the position of the screens. Furthermore, the disclosure of the two Podoleanu's papers was restricted to the simplification of the spectral terms encountered in the Fourier transform of the channeled spectrum when a cavity low coherence source was used.